Linux 4.14 compat: IO acct, global_page_state, etc (openzfs#6655)

generic_start_io_acct/generic_end_io_acct in the master branch of the linux kernel requires that the request_queue be provided. Move the logic from freemem in the spl to arc_free_memory in arc.c. Do this so we can take advantage of global_page_state interface checks in zfs. Upstream kernel replaced struct block_device with struct gendisk in struct bio. Determine if the function bio_set_dev exists during configure and have zfs use that if it exists. bio_set_dev torvalds/linux@74d4699 global_node_page_state torvalds/linux@75ef718 io acct torvalds/linux@d62e26b Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Giuseppe Di Natale <dinatale2@llnl.gov> Closes openzfs#6635 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
tonyhutter · Sep 19, 2017 · 661907e · 661907e
1 parent d3e7d98
commit 661907e
Show file tree

Hide file tree

Showing 7 changed files with 112 additions and 19 deletions.
diff --git a/config/kernel-bio_set_dev.m4 b/config/kernel-bio_set_dev.m4
@@ -0,0 +1,22 @@
+dnl #
+dnl # Linux 4.14 API,
+dnl #
+dnl # The bio_set_dev() helper was introduced as part of the transition
+dnl # to have struct gendisk in struct bio. 
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_BIO_SET_DEV], [
+	AC_MSG_CHECKING([whether bio_set_dev() exists])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/bio.h>
+		#include <linux/fs.h>
+	],[
+		struct block_device *bdev = NULL;
+		struct bio *bio = NULL;
+		bio_set_dev(bio, bdev);
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_BIO_SET_DEV, 1, [bio_set_dev() exists])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
diff --git a/config/kernel-generic_io_acct.m4 b/config/kernel-generic_io_acct.m4
@@ -4,8 +4,8 @@ dnl #
 dnl # torvalds/linux@394ffa503bc40e32d7f54a9b817264e81ce131b4 allows us to
 dnl # increment iostat counters without generic_make_request().
 dnl #
-AC_DEFUN([ZFS_AC_KERNEL_GENERIC_IO_ACCT], [
-	AC_MSG_CHECKING([whether generic IO accounting symbols are avaliable])
+AC_DEFUN([ZFS_AC_KERNEL_GENERIC_IO_ACCT_3ARG], [
+	AC_MSG_CHECKING([whether 3 arg generic IO accounting symbols are available])
 	ZFS_LINUX_TRY_COMPILE_SYMBOL([
 		#include <linux/bio.h>
 
@@ -18,8 +18,35 @@ AC_DEFUN([ZFS_AC_KERNEL_GENERIC_IO_ACCT], [
 		generic_end_io_acct(0, NULL, 0);
 	], [generic_start_io_acct], [block/bio.c], [
 		AC_MSG_RESULT(yes)
-		AC_DEFINE(HAVE_GENERIC_IO_ACCT, 1,
-		    [generic_start_io_acct()/generic_end_io_acct() avaliable])
+		AC_DEFINE(HAVE_GENERIC_IO_ACCT_3ARG, 1,
+		    [generic_start_io_acct()/generic_end_io_acct() available])
+	], [
+		AC_MSG_RESULT(no)
+	])
+])
+
+dnl #
+dnl # Linux 4.14 API,
+dnl #
+dnl # generic_start_io_acct/generic_end_io_acct now require request_queue to be
+dnl # provided. No functional changes, but preparation for inflight accounting 
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_GENERIC_IO_ACCT_4ARG], [
+	AC_MSG_CHECKING([whether 4 arg generic IO accounting symbols are available])
+	ZFS_LINUX_TRY_COMPILE_SYMBOL([
+		#include <linux/bio.h>
+
+		void (*generic_start_io_acct_f)(struct request_queue *, int,
+		    unsigned long, struct hd_struct *) = &generic_start_io_acct;
+		void (*generic_end_io_acct_f)(struct request_queue *, int,
+		    struct hd_struct *, unsigned long) = &generic_end_io_acct;
+	], [
+		generic_start_io_acct(NULL, 0, 0, NULL);
+		generic_end_io_acct(NULL, 0, NULL, 0);
+	], [generic_start_io_acct], [block/bio.c], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_GENERIC_IO_ACCT_4ARG, 1,
+		    [generic_start_io_acct()/generic_end_io_acct() 4 arg available])
 	], [
 		AC_MSG_RESULT(no)
 	])

diff --git a/config/kernel.m4 b/config/kernel.m4
@@ -24,6 +24,7 @@ AC_DEFUN([ZFS_AC_CONFIG_KERNEL], [
 	ZFS_AC_KERNEL_BDEV_PHYSICAL_BLOCK_SIZE
 	ZFS_AC_KERNEL_BIO_BVEC_ITER
 	ZFS_AC_KERNEL_BIO_FAILFAST_DTD
+	ZFS_AC_KERNEL_BIO_SET_DEV
 	ZFS_AC_KERNEL_REQ_FAILFAST_MASK
 	ZFS_AC_KERNEL_REQ_OP_DISCARD
 	ZFS_AC_KERNEL_REQ_OP_SECURE_ERASE
@@ -113,7 +114,8 @@ AC_DEFUN([ZFS_AC_CONFIG_KERNEL], [
 	ZFS_AC_KERNEL_KMAP_ATOMIC_ARGS
 	ZFS_AC_KERNEL_FOLLOW_DOWN_ONE
 	ZFS_AC_KERNEL_MAKE_REQUEST_FN
-	ZFS_AC_KERNEL_GENERIC_IO_ACCT
+	ZFS_AC_KERNEL_GENERIC_IO_ACCT_3ARG
+	ZFS_AC_KERNEL_GENERIC_IO_ACCT_4ARG
 	ZFS_AC_KERNEL_FPU
 	ZFS_AC_KERNEL_KUID_HELPERS
 	ZFS_AC_KERNEL_MODULE_PARAM_CALL_CONST

diff --git a/include/linux/blkdev_compat.h b/include/linux/blkdev_compat.h
@@ -598,16 +598,26 @@ blk_queue_discard_granularity(struct request_queue *q, unsigned int dg)
  */
 #define	VDEV_HOLDER			((void *)0x2401de7)
 
-#ifndef HAVE_GENERIC_IO_ACCT
 static inline void
-generic_start_io_acct(int rw, unsigned long sectors, struct hd_struct *part)
+blk_generic_start_io_acct(struct request_queue *q, int rw,
+    unsigned long sectors, struct hd_struct *part)
 {
+#if defined(HAVE_GENERIC_IO_ACCT_3ARG)
+	generic_start_io_acct(rw, sectors, part);
+#elif defined(HAVE_GENERIC_IO_ACCT_4ARG)
+	generic_start_io_acct(q, rw, sectors, part);
+#endif
 }
 
 static inline void
-generic_end_io_acct(int rw, struct hd_struct *part, unsigned long start_time)
+blk_generic_end_io_acct(struct request_queue *q, int rw,
+    struct hd_struct *part, unsigned long start_time)
 {
-}
+#if defined(HAVE_GENERIC_IO_ACCT_3ARG)
+	generic_end_io_acct(rw, part, start_time);
+#elif defined(HAVE_GENERIC_IO_ACCT_4ARG)
+	generic_end_io_acct(q, rw, part, start_time);
 #endif
+}
 
 #endif /* _ZFS_BLKDEV_H */
diff --git a/module/zfs/arc.c b/module/zfs/arc.c
@@ -3988,6 +3988,24 @@ arc_all_memory(void)
 #endif
 }
 
+#ifdef _KERNEL
+static uint64_t
+arc_free_memory(void)
+{
+#ifdef ZFS_GLOBAL_NODE_PAGE_STATE
+	return (nr_free_pages() +
+	    global_node_page_state(NR_INACTIVE_FILE) +
+	    global_node_page_state(NR_INACTIVE_ANON) +
+	    global_node_page_state(NR_SLAB_RECLAIMABLE));
+#else
+	return (nr_free_pages() +
+	    global_page_state(NR_INACTIVE_FILE) +
+	    global_page_state(NR_INACTIVE_ANON) +
+	    global_page_state(NR_SLAB_RECLAIMABLE));
+#endif
+}
+#endif
+
 typedef enum free_memory_reason_t {
 	FMR_UNKNOWN,
 	FMR_NEEDFREE,
@@ -4024,7 +4042,7 @@ arc_available_memory(void)
 	int64_t lowest = INT64_MAX;
 	free_memory_reason_t r = FMR_UNKNOWN;
 #ifdef _KERNEL
-	uint64_t available_memory = ptob(freemem);
+	uint64_t available_memory = ptob(arc_free_memory());
 	int64_t n;
 #ifdef __linux__
 	pgcnt_t needfree = btop(arc_need_free);
@@ -6059,7 +6077,7 @@ static int
 arc_memory_throttle(uint64_t reserve, uint64_t txg)
 {
 #ifdef _KERNEL
-	uint64_t available_memory = ptob(freemem);
+	uint64_t available_memory = ptob(arc_free_memory());
 	static uint64_t page_load = 0;
 	static uint64_t last_txg = 0;
 #ifdef __linux__

diff --git a/module/zfs/vdev_disk.c b/module/zfs/vdev_disk.c
@@ -498,6 +498,14 @@ vdev_submit_bio_impl(struct bio *bio)
 #endif
 }
 
+#ifndef HAVE_BIO_SET_DEV
+static inline void
+bio_set_dev(struct bio *bio, struct block_device *bdev)
+{
+	bio->bi_bdev = bdev;
+}
+#endif /* !HAVE_BIO_SET_DEV */
+
 static inline void
 vdev_submit_bio(struct bio *bio)
 {
@@ -580,7 +588,7 @@ __vdev_disk_physio(struct block_device *bdev, zio_t *zio,
 		/* Matching put called by vdev_disk_physio_completion */
 		vdev_disk_dio_get(dr);
 
-		dr->dr_bio[i]->bi_bdev = bdev;
+		bio_set_dev(dr->dr_bio[i], bdev);
 		BIO_BI_SECTOR(dr->dr_bio[i]) = bio_offset >> 9;
 		dr->dr_bio[i]->bi_end_io = vdev_disk_physio_completion;
 		dr->dr_bio[i]->bi_private = dr;
@@ -654,7 +662,7 @@ vdev_disk_io_flush(struct block_device *bdev, zio_t *zio)
 
 	bio->bi_end_io = vdev_disk_io_flush_completion;
 	bio->bi_private = zio;
-	bio->bi_bdev = bdev;
+	bio_set_dev(bio, bdev);
 	bio_set_flush(bio);
 	vdev_submit_bio(bio);
 	invalidate_bdev(bdev);

diff --git a/module/zfs/zvol.c b/module/zfs/zvol.c
@@ -761,7 +761,8 @@ zvol_write(void *arg)
 	ASSERT(zv && zv->zv_open_count > 0);
 
 	start_jif = jiffies;
-	generic_start_io_acct(WRITE, bio_sectors(bio), &zv->zv_disk->part0);
+	blk_generic_start_io_acct(zv->zv_queue, WRITE, bio_sectors(bio),
+	    &zv->zv_disk->part0);
 
 	sync = bio_is_fua(bio) || zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS;
 
@@ -794,7 +795,8 @@ zvol_write(void *arg)
 		zil_commit(zv->zv_zilog, ZVOL_OBJ);
 
 	rw_exit(&zv->zv_suspend_lock);
-	generic_end_io_acct(WRITE, &zv->zv_disk->part0, start_jif);
+	blk_generic_end_io_acct(zv->zv_queue, WRITE, &zv->zv_disk->part0,
+	    start_jif);
 	BIO_END_IO(bio, -error);
 	kmem_free(zvr, sizeof (zv_request_t));
 }
@@ -840,7 +842,8 @@ zvol_discard(void *arg)
 	ASSERT(zv && zv->zv_open_count > 0);
 
 	start_jif = jiffies;
-	generic_start_io_acct(WRITE, bio_sectors(bio), &zv->zv_disk->part0);
+	blk_generic_start_io_acct(zv->zv_queue, WRITE, bio_sectors(bio),
+	    &zv->zv_disk->part0);
 
 	sync = bio_is_fua(bio) || zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS;
 
@@ -881,7 +884,8 @@ zvol_discard(void *arg)
 		zil_commit(zv->zv_zilog, ZVOL_OBJ);
 
 	rw_exit(&zv->zv_suspend_lock);
-	generic_end_io_acct(WRITE, &zv->zv_disk->part0, start_jif);
+	blk_generic_end_io_acct(zv->zv_queue, WRITE, &zv->zv_disk->part0,
+	    start_jif);
 	BIO_END_IO(bio, -error);
 	kmem_free(zvr, sizeof (zv_request_t));
 }
@@ -902,7 +906,8 @@ zvol_read(void *arg)
 	ASSERT(zv && zv->zv_open_count > 0);
 
 	start_jif = jiffies;
-	generic_start_io_acct(READ, bio_sectors(bio), &zv->zv_disk->part0);
+	blk_generic_start_io_acct(zv->zv_queue, READ, bio_sectors(bio),
+	    &zv->zv_disk->part0);
 
 	while (uio.uio_resid > 0 && uio.uio_loffset < volsize) {
 		uint64_t bytes = MIN(uio.uio_resid, DMU_MAX_ACCESS >> 1);
@@ -922,7 +927,8 @@ zvol_read(void *arg)
 	zfs_range_unlock(zvr->rl);
 
 	rw_exit(&zv->zv_suspend_lock);
-	generic_end_io_acct(READ, &zv->zv_disk->part0, start_jif);
+	blk_generic_end_io_acct(zv->zv_queue, READ, &zv->zv_disk->part0,
+	    start_jif);
 	BIO_END_IO(bio, -error);
 	kmem_free(zvr, sizeof (zv_request_t));
 }